FFmpeg heuristic detector (--detector ffmpeg)¶

Fastest detector. Uses FFmpeg's astats filter to get RMS/peak energy statistics per segment.

Usage¶

praisonai-editor edit file.mp3 --preset no_silence --detector ffmpeg

How it works¶

flowchart LR
    A[Audio] --> B[ffmpeg astats]
    B --> C[RMS dB per 0.5s window]
    C --> D{Below silence\nthreshold?}
    D -->|Yes| E[silence]
    D -->|No| F[speech or music\nbased on energy]

When to use ffmpeg¶

• Very long files where speed matters
• Simple silence removal only (--preset no_silence)
• When you don't need precise speech/music classification
• Final step in a pipeline after rough cutting

Limitations¶

• Cannot distinguish speech from music
• No machine learning — accuracy is heuristic only